macos-computer-use

Automates macOS GUI interactions via accessibility APIs and screenshot capture.

Updated Jul 6, 2026
One-click install
npx skills add https://github.com/Lucien-1127/strata-skill --skill macos-computer-use-lucien-1127
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: macos-computer-use
Source: https://github.com/Lucien-1127/strata-skill/tree/main/macos-computer-use
Command: npx skills add https://github.com/Lucien-1127/strata-skill --skill macos-computer-use-lucien-1127

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Automating native macOS applications is difficult when tasks require clicking, typing, scrolling, and reading screen content in apps that have no API, and coordinate-based automation is fragile and steals the user's focus. ## Core Features & Use Cases - Element-indexed interaction: Capture screens with numbered overlays on every interactable element, then click, drag, or scroll by element index instead of pixel coordinates. - Background operation: Drive apps on any Space without moving the user's cursor, stealing keyboard focus, or raising windows. - Model-agnostic workflow: Works with any tool-capable model (Claude, GPT, Gemini, or local models) through a single computer_use tool with capture modes for vision or text-only models. - Use Case: Ask the agent to open Safari, navigate to a site, fill in a form, and verify the result — all while you keep working in another window. ## Quick Start Use the computer_use tool to capture the Safari window and click the Sign In button by its element index.

Frequently Asked Questions about macos-computer-use

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate clicking and typing in macOS apps with an AI agent?▼

Use the computer_use tool: first call capture with mode=som to get a screenshot with numbered element overlays, then call click or type referencing the element index. Re-capture after state-changing actions to verify results.

What is the difference between element index and coordinate clicking in GUI automation?▼

Element indices come from the accessibility tree and remain reliable across models, while pixel coordinates are fragile when UI layouts shift. The skill recommends clicking by element index and re-capturing whenever the UI changes.

Does macOS computer use work with models other than Claude?▼

Yes, the workflow is model-agnostic and works with any tool-capable model including GPT, Gemini, and local models via OpenAI-compatible endpoints. There is no Anthropic-specific schema required.

Can GUI automation run in the background without stealing focus on macOS?▼

Yes, cua-driver routes input to apps without raising windows, moving the cursor, or switching Spaces. You should scope captures to a specific app and avoid raise_window unless the user explicitly requests it.

When should I use browser tools instead of macOS computer use?▼

Use browser_* tools for web automation since they drive a headless Chromium more reliably. Reserve computer_use for tasks requiring the user's actual native Mac apps like Mail, Messages, Finder, or Figma.

Why does computer_use fail with 'cua-driver not installed'?▼

The cua-driver dependency is missing. Run hermes tools and enable Computer Use, which installs cua-driver via its upstream script; macOS Accessibility and Screen Recording permissions are also required.